Design to end-of-life margins, or design twice.
A span that closes on day one and fails in year eight was never designed. It was measured.
Optical budgets age. Connectors degrade with every mating cycle, splices accumulate through repairs, and fibre attenuation rises as an estate is disturbed by adjacent works. A budget that closes against beginning-of-life numbers tells you the network works today. It says nothing about year eight of a 20-year concession.
Our practice is to budget every span event by event, then prove it against the worst-case, end-of-life limit: repair margin, ageing margin and the coldest OSNR the receiver will ever see. If the margin does not close, the design changes before the civils do. Moving a chamber on a drawing costs minutes. Moving one under a London street costs a traffic order.
This is also where vendor pre-sales studies quietly diverge from engineering. A free design exists to make a product line close the case, and the easiest way to close a case is to assume the kindest conditions. An independent budget assumes the unkindest, because that is the network the operator will actually own.
End-of-life discipline changes behaviour upstream too. Reach budgets from Base Station Hotels stop being a single number and become a policy: which spans may take a future splice, which laterals can be extended, where headroom is reserved for services that do not exist yet, CBTC and ESN among them.
The test we apply is simple to state. Would you sign the margin in front of an assurance panel, knowing the panel will still exist in a decade? If the answer needs caveats, the design is not finished.
First-time approval is not luck. It is the visible end of a chain that starts with refusing to design to the demo-day case.